What is red teaming?

Deliberately testing an AI system in challenging or unexpected ways to find harmful behavior or weak protections.

Also called Red team, AI red-teaming.

For example: Testers try to make an assistant reveal information it should keep private.